8 March 2017
This lab uses statistical software program called CrimeStat to analyse the spatial distribution of Auto thefts (AT), Commercial break and enter events (Commercial B&E) and Residential break and enter events (Residential B&E) that occurred in Nepean, Ottawa, Ontario between January 2005 and March 2006. Results of the analysis conclude that the crime events in Nepean are clustered and not randomly distributed. This conclusion remained true even when the distribution of population was taken into account.
This was my favourite lab out of all the labs that we did this semester. I really enjoyed learning about all the different tools and statistics that can be used to conduct spatial analyses. Below is a list of these tools and statistics (and it is a long list, holy smokes!).
- Using spatial indices to determine spatial clustering
- Nearest neighbour analysis
- Moran’s I analysis + what a correlogram is
- Using spatial clustering techniques to identify spatial clusters
- Point locations – fuzzy mode clustering routine
- Nearest neighbour hierarchical spatial clustering routine
- Nearest neighbour hierarchical spatial clustering routine – risk adjusted
- Using the Knox index to conduct a time-space analysis
- Using kernel density interpolation to visualise spatial clustering
- Single kernel density interpolation
- Dual kernel density interpolation
If I had to choose just one tool to use to visualise the crime data in this study, I would choose the kernel density surfaces. This is because the kernel density technique produces a continuous density surface which in turn allows users to identify hot spots themselves. In comparison, clustering analyses such as Fuzzy Mode and Nearest Neighbour Hierarchical (NNH) displays crime frequencies as point data and polygons, respectively which ends up dictating the user’s identification of hot spots (since hot spots are identified by the routines themselves).
Click here to see the map I created to compare the the results of the single and dual kernel densities. As shown, changes in density of crime in the region are displayed over a continuous surface which allows users to identify hot spots themselves.
Click here to see the map I created to display the results of the the Fuzzy Mode clustering routine. As shown, crime frequencies are displayed as points. While this also display densities, the user’s identification of hot spots is solely based on the hot spots identified by the routines (NIJ, 2004).
Click here to see the map I created to display the results of the the NNH routine. As shown, crime frequencies are displayed as polygons. Similar to the Fuzzy Mode map above, user’s identification of hot spots is based on the polygons identified by the routine.
References
The National Institute of Justice. 2004. Chapter 7: Kernel Density Interpolation. CrimeStat III A spatial statistics program for the analysis of crime incident locations. Retrieved from https://www.icpsr.umich.edu/CrimeStat/files/CrimeStatChapter.8.pdf